Skip to content

Graph Insights Tutorial – Supply Chain Risk Analysis¤

This tutorial demonstrates a workflow of Graph Insights using the Northwind dataset.

Demo Data

The demo data using in this tutorial is available here: northwind-data.ttl. You can import the data:

  • by using the web interface (Explore -> Knowledge graphs → Add new graph) or

  • by using the command line interface:

    cmemc graph import northwind-data.ttl http://example.org/northwind/

Business Scenario¤

Northwind Traders needs to identify vulnerabilities in the trade network by answering a critical question:

“Which VIP US customers will be impacted if the UK supply chain fails, and which products will be involved?”

We will traverse the ordering chain, identify high-volume US partners, and isolate the product mix.

1. Seed the Canvas¤

Every exploration requires a starting point. We begin with the central entity: The product.

  1. Locate the Product class in the class tree in the left pane.
  2. Drag it to the center of the canvas.
  3. Outcome: A group appears, representing the entire inventory (e.g., 77 Products).

Start with Product Category

2. Build Traversals to the Interest Groups¤

Next, we expand the exploration tree to connect Suppliers and Customers.

Preparation¤

To simplify the expansion menus, hide generic classes that are not relevant to this analysis. - Right-click the Agent class in the class tree. - Select Deactivate class.

Expansion Steps¤

We extend the exploration tree from the central Product group.

  1. Add suppliers:

    • Select the Product group.
    • Expand to the Supplier class via → hasSupplier.

  2. Add customers:

    • Select the Product group again.
    • Expand to the Order class via ← includesProduct.
    • From the new Order group, expand to the Customer class via → hasCustomer.

Supply Chain Expansion

3. Filter Key Players¤

The exploration tree currently visualizes all connections. We must now isolate the high-value targets (VIPs) and the specific risk region (UK).

3.1. Identify VIP US Customers¤

We define “VIP” as customers with the highest volume of incoming orders.

  1. Select the Customer group and open the resource table.
  2. Filter by country: Open the menu of the country column from its header (using the menu dropdown or a right-click) and enter USA in the text field of the filter submenu.
  3. Identify VIPs: Click the Predecessors column header to Sort Descending. This ranks customers by their incoming connection count (number of Orders).

USA Customer Filter

  1. Select and restrict: Check the boxes for the top 5 customers and click Restrict to selection.

Top USA Customer Filter

3.2. Isolate UK Suppliers¤

  1. Select the Supplier group to open the table.
  2. Filter Country to UK.
  3. Click Apply filters.

UK Supplier Filter

The Disconnected State

At this stage, you have filtered the ends of both traversals, but their start (Products) still shows the total inventory (left-join pattern). The branches of the exploration tree represent independent aggregated traversals sharing the same starting point.

Supply Chain without Backpropagation

4. Propagate the Filtering to the Products¤

We now need to identify the intersection: Which products are both sold by UK Suppliers and bought by VIP US Customers?

We apply left group restriction (i.e., switching from left join to inner join) on the connections to retain only resources of the left group having at least one object property assertion in the connection (i.e., successor in the right group).

4.1. Identify Products from UK Suppliers¤

Right-click the connection beam between Supplier and Product and select Restrict left group:

  • This retains only products having a UK supplier.
  • As this reduces the start group of the second branch of the exploration tree, the Order and Customer groups might be impacted.

4.2. Identify Products ordered by VIP USA Customers¤

Apply the same restriction to the connections between Order and Customer as well as between Product and Order:

  • This retains only products appearing in at least one order of at least one VIP USA customer.
  • All products having been constrained to have a UK supplier in the previous step, the Products group now contains the high risk products.

Supply Chain with Backpropagation

5. Analyze Impact¤

We have isolated the high risk product group, that we can now analyze:

  1. Select the filtered Product group to display the class histogram in the right pane.
  2. Expand the Product class bar to display its sub-classes (e.g., Beverages, Confections).
  3. Insight: The products of the Confections class are the most represented in this high impact risk group.

Product Histogram for Supply Chain

6. Operationalize Findings¤

Finally, we convert this visual insight into an actionable list for the procurement team.

  1. Drill down:

    • In the histogram, right-click the Confections bar.
    • Select Restrict group to class. This discards the other product types, leaving only the high-risk confections, with the side effect of reducing the UK suppliers to a single resource.

  2. Visualize traversals including a specific resource:

    • Open the resource table of the Products group.
    • Right-click a specific high-value item (e.g., “Teatime Chocolate Biscuits”) to open its context menu.
    • Select Flag all connected resources on the branch.

Supply Chain: Flag Branch

  • Graph Insights highlights all resources on a specific high-risk traversal: The specific UK Supplier → The specific Biscuit → The specific VIP US Customers buying it.

Supply Chain: Flagged Branch

Summary and Next Steps¤

You have successfully modeled a complex dependency network, filtered it by geo-location and volume, and identified a risk cluster.

Comments